Predicting the Effect of Hardware Fault Injection

Abstract

Fault attacks on the hardware are now common to extract valuable information from a device. The most used means is the electromagnetic perturbation. Designers have to verify how robust are their designs against this kind of attack. We designed a prediction tool that generates mutants according to a fault model and then checks security properties on the mutants. We improved the generation of mutants at the instruction level and at the function level to reduce the evaluation time. we reduce that number between 32% and 84% according to the fault model.